Negotiation and Bargaining

Agents negotiate resources, prices, task allocations, and terms. The 2026 benchmark set is clear: NegotiationArena (arXiv:2402.05863) shows LLMs can improve payoffs ~20% via persona manipulation ("desperation"); "Measuring Bargaining Abilities" (arXiv:2402.15813) shows buyer is harder than seller and scale does not help — their **OG-Narrator** (deterministic offer generator + LLM narrator) pushed deal rate from 26.67% to 88.88%; the Large-Scale Autonomous Negotiation Competition (arXiv:2503.06416) ran ~180k negotiations and found that **chain-of-thought-concealing** agents win by hiding reasoning from counterparts; Bhattacharya et al. 2025 on Harvard Negotiation Project metrics ranked Llama-3 most-effective, Claude-3 aggressive, GPT-4 fairest. This lesson implements Contract Net Protocol (the FIPA ancestor, Lesson 02), wires an LLM-style buyer/seller, runs an OG-Narrator-style decomposition, and measures how deal rate changes with each structural choice.

Learn + Build 1.3 hr Python (stdlib)

Prerequisites: Phase 16 · 02 (FIPA-ACL Heritage), Phase 16 · 09 (Parallel Swarm Networks)

#Problem

Two agents need to agree on a price. Left to themselves with pure language prompts, 2024-2026 LLMs close deals at surprisingly low rates (~27% on tightly-parameterized bargains in arXiv:2402.15813). Scale does not fix it: GPT-4 is not structurally better at bargaining than GPT-3.5; it is better at the language of bargaining.

The root issue is that LLMs conflate two jobs — deciding the offer and narrating the offer. OG-Narrator separated these: a deterministic offer generator computes numeric moves; the LLM only narrates. Deal rate jumps to ~89%.

This mirrors a classical multi-agent finding: decoupling the mechanism from the communication layer wins. Contract Net Protocol (FIPA, 1996; Smith, 1980) is the reference task-market mechanism. Plug an LLM into the narration slot and you get a modern LLM-powered task market.

#Problem

Keep going. The rest of this lesson is one step away.